fix: make cpufeature a local import

setup.py

@@ -33,7 +33,6 @@ import torch
 import torch.version
 from wheel.bdist_wheel import bdist_wheel as _bdist_wheel
 from setuptools import setup, Extension
-from cpufeature.extension import CPUFeature
 from torch.utils.cpp_extension import BuildExtension, CUDAExtension, CUDA_HOME, ROCM_HOME
 try:
     from torch_musa.utils.simple_porting import SimplePorting
@@ -188,6 +187,8 @@ class VersionInfo:
             raise ValueError(
                 "Unsupported cpu Instructions: {}".format(flags_line))
         elif sys.platform == "win32":
+            from cpufeature.extension import CPUFeature
+
             if CPUFeature.get("AVX512bw", False):
                 return 'fancy'
             if CPUFeature.get("AVX512f", False):